Best of every.to on Kiwi News
Most upvoted by Kiwi News readers
13 stories from every.to have been shared on Kiwi News since 3 July 2023, with 31 upvotes and 2 comments in total. The latest was shared on 23 May 2025. They are listed below, most upvoted first.
Text under a title is an AI summary of the linked article.
- Content moderation double-bind theory
The article argues that content moderation creates a double-bind for every company that hosts or publishes user content: the closer content gets to the line of what is prohibited, the more attention it attracts, and the more attention it attracts, the costlier it is to moderate. Citing a 2018 Facebook note by Mark Zuckerberg, the author says engagement rises as content nears the policy line, even when users say afterward that they dislike it. The author gives Logan Paul, David Dobrik, and Basecamp as examples, and argues that ad-based revenue is not the root cause, since subscription platforms such as Substack also reward edgy content. The article also contends that Section 230 lets platforms present themselves as tools rather than publishers.
- The Knowledge Economy Is Over. Welcome to the Allocation Economy
Dan Shipper argues that the knowledge economy is giving way to an "allocation economy," in which the most valuable skill is managing resources, including AI models, to get work done rather than knowing things yourself. He says people will move from doing the work to choosing which work gets done, judging whether it is good enough, and editing it, much as managers do today. Shipper notes that only about 1 million managers in the U.S., roughly 12% of the workforce, currently practice these skills, and that cheap AI could let everyone become a "model manager." He also cites Tyler Cowen's 2013 book Average Is Over, which predicted a split between workers whose skills complement computers and those left behind, and says the shift will be gradual.
- Why You Should Build in Deep Tech
Venture capitalist Leo Polovets argues that deep tech is currently the best sector to invest in and build companies in, citing PitchBook data that he says contradicts three common misconceptions. He says deep tech companies are not more capital intensive than traditional tech, with the median company that exited above $250 million having raised 11-15% of its exit size in prior funding. He also reports that deep tech and life science companies exit 10-25% faster than traditional counterparts, and that defense, space tech and life science companies have 2-5% odds of a $250 million-plus exit, compared with 1-1.5% for SaaS, fintech and AI/ML.
- How to Price Generative AI Products
Ian Clark argues that generative AI startups should price their products based on customers' willingness to pay rather than on their costs or competitors' prices, even though large language models add marginal costs. He says cost has no bearing on the optimal price in his lemonade example, and that costs only matter once gross margins fall to 60% or lower. He recommends at least 10 customer interviews per month, choosing a price metric that correlates with willingness to pay, and using a Van Westendorp price study, which was introduced by Dutch economist Peter van Westerndorp in 1976.
- What happened to the creator economy?
The author argues that the creator economy startup boom has collapsed because investors misread the risks, treating creator tools as easy vertical SaaS bets in a market where the top 1% of creators earn most of the money. Creator economy funding fell 86% to $123M, and Patreon, Linktree, Cameo and Substack all made layoffs. The author says the startups that scaled, such as OnlyFans, Substack, Cameo and Kajabi, did so by serving needs that large platforms ignored, while companies building general tools faced tough competition from established products like Quickbooks.
- The Knowledge Economy Is Over. Welcome to the Allocation Economy
Dan Shipper argues that the knowledge economy is giving way to an "allocation economy," in which people are judged less on what they know and more on how well they direct and manage work, including work done by AI models. He points to his own use of ChatGPT for summarizing, which led him to hand that task to AI and focus on editing and directing it instead. Shipper says most workers will become "model managers" who need skills such as a coherent vision, clear taste, evaluating models, and knowing when to get into the details. He describes the shift as gradual and calls it a good thing, while noting that it could widen the gap between people who can use language models well and those who cannot.
- Vibe Check: Claude 4 Opus
Dan Shipper reviews Anthropic's Claude 4 Opus, concluding that it is a strong model for editing, coding and long research tasks, though he still uses OpenAI's o3 for everyday work. He says Opus edits honestly, judging writing against a set of principles he and Danny Aziz provided, and it tracked those principles across long prompts where other reasoning models lost them. In Claude Code it built a feature for Cora that was close to shippable, and its research mode spawned parallel agents that gathered 645 sources for one query. Shipper adds that model performance is now good enough that user interfaces and features like memory matter more, and that ChatGPT's memory keeps him on it.
- Asset-light Software Businesses
Will Manidis and John Kennedy argue that software businesses are shifting from fixed costs to variable costs, producing asset-light companies that founders can build and fund with more control over their futures. They say the median bootstrapped B2B SaaS company spends around 90% of its annual recurring revenue on costs, including 25% on go-to-market and 24% on research and development. The authors contend that generative AI will accelerate this change by letting companies automate junior-level white-collar work, and they compare it to the internet's effect on content distribution. The article also states that Github Copilot has $100 million in average recurring revenue but loses around $20 per month per user.
- Venture Capital Is Ripe for Disruption
The article argues that venture capital's focus on rare, very large outcomes is failing most investors and founders, and it sees an opening for funds aimed at smaller but more likely successes. It cites Figma's $20B acquisition: Index Ventures owned about 13%, worth roughly $2.6B, yet the author says that stake would not return Index's $3.1B fund, which was raised in 2021. The piece also says only 48 public tech companies are valued above $50B while close to 1,000 venture funds compete for such outcomes, and that Black entrepreneurs received 0.12% of venture dollars in Q3 of that year. It proposes funds of $150M or less that offer equity optionality, though it does not present this as tested.
- The product wedge: How to scope your initial product
The article argues that startups stuck in customer acquisition should sharpen their initial product into a "wedge" rather than pivot, meaning a early feature that is easy for customers to adopt and feasible for the company to deliver before it has scale advantages. It defines the approach by two principles: some value propositions are easier for customers to adopt, and some are easier for new companies to deliver. It cites TikTok's start as a music-video recording app, where users "came for the tool, but then ended up staying for the network," and says Chris Dixon popularized the "thin edge of the wedge" strategy in 2010. The piece presents five types of product wedges, with examples including Instagram, Glossier and Netflix.
- Tokenomics 103: Utility
This third part of a tokenomics series argues that a token needs real utility to create demand, and it frames utility around four questions: whether a token is meant to be spent or held, whether it generates cash flows, whether it confers governance rights, and whether it can be used as collateral. The article uses Chainlink as an example of a spending token: LINK is primarily used to pay for Chainlink services, and it had lost 50-90% of its value against ETH for anyone who bought after July 2019 and held. Among cash-flow examples, it says Convex was paying a 44% APR on locked Convex, while SUSHI stakers receiving xSUSHI earned fees at about 10% APR. It also notes that AAVE governance votes often attract 250,000 to 400,000 votes.
- Is it a Ponzi?
The article proposes a framework for judging crypto projects by whether they are living or dying, based on whether users pay for a service more than the project pays out in incentives. It describes living projects such as AAVE, Uniswap and Yearn, where users pay fees without receiving tokens, and dying projects such as Terra's Anchor Protocol, whose emissions exceeded its revenue. It sorts projects into four further categories: pyramid schemes, Ponzis, tulips and spenders. Bitconnect is cited as the most famous crypto Ponzi, while the article says Anchor was not clearly a Ponzi because it did not appear to commit fraud.
- Tokenomics 101: The basics of evaluating cryptocurrencies
The article argues that a crypto token's value depends on its tokenomics, meaning the supply mechanics and the demand drivers that shape holder behavior, and that well-designed tokenomics make long-term success more likely. It says supply is assessed through total and circulating amounts, emission rate, and allocation, with Bitcoin's 21,000,000 cap cited as an example of limited inflation. On demand, it points to return on investment from staking or protocol revenue, community belief ("memes"), and lockup incentives, using Curve's up-to-four-year locking rewards and Convex Finance's fixed 100m supply as examples.